More than half of Australian enterprises say their expertise structure is just too inflexible, too pricey and too sluggish to assist the AI methods they’re now attempting to construct.
Research overlaying 200 native organizations, carried out by IDC and MongoDB, discovered that 58% described their current infrastructure that approach — a verdict on the methods companies have been working properly earlier than agentic AI arrived.
The discovering lands as a world reckoning takes maintain over what agentic AI truly calls for from enterprise IT. Systems constructed to deal with predictable, human-triggered software program requests at the moment are being requested to assist AI brokers that autonomously name fashions, question databases, and coordinate with different brokers throughout a number of steps. That’s a job description most company expertise estates have been by no means designed to fill.
Google Cloud places a quantity on the worldwide hole
Google Cloud’s 2026 State of AI Infrastructure report, printed July eight based mostly on a survey of 1,402 IT leaders worldwide, discovered that 83% of organizations want infrastructure upgrades earlier than agentic AI can run in manufacturing somewhat than in a pilot. The report traces the hole to rising inference prices, operational complexity, governance calls for, and energy constraints as brokers tackle multistep work throughout enterprise methods.
52% of respondents now run hybrid multicloud architectures, and 48% prioritize infrastructure with strict data-residency controls. Both are indicators {that a} single cloud area or a single vendor’s stack is not handled as ample. Edge deployment has additionally grow to be a mainstream requirement somewhat than a distinct segment one: 90% known as it necessary to their AI plans, with 72% ranking it extraordinarily or essential.
Cost is the clearest symptom of the mismatch. 62% of respondents reported a big “inference tax” — data-egress charges, extra storage, idle specialised {hardware} — that has nothing to do with mannequin licensing and every thing to do with infrastructure constructed for a unique workload. Another 81% cited operational complexity itself as a hidden value of scaling brokers.
Why outdated structure and new brokers do not match
The core battle is structural, not beauty. Traditional enterprise methods have been designed round discrete, stateless transactions: a person submits a request, a system responds, and the interplay ends.
Agentic AI wants methods that retain context throughout long-running workflows, recuperate cleanly from failure partway by means of a process, and observe what an autonomous course of did and why, typically with no human confirming every step.
That mismatch is displaying up in Australian governance knowledge too. Deloitte’s State of AI within the Enterprise analysis discovered solely 12% of Australian leaders say generative AI is already reworking their enterprise, in contrast with 25% globally. That performs out whilst 57% of Australian organizations report deploying robotics, whereas curiosity in sovereign, domestically accountable AI infrastructure continues to climb.
Deloitte discovered 72% of Australian corporations now weigh a vendor’s nation of origin in AI selections, and greater than 80% deal with sovereignty as a strategic planning precedence.
Power and knowledge placement selections are converging in consequence. The federal authorities’s March 2026 expectations for knowledge facilities and AI infrastructure builders explicitly hyperlink new “AI factory” builds to nationwide safety and knowledge sovereignty, not simply capability.
Google’s world survey discovered the same sample on the customer aspect: 91% of IT leaders now issue energy consumption into {hardware} choice, and 61% fee it as a main or vital issue.
Governance is the opposite half of the readiness hole
Security and governance are the place the old-versus-new battle is most acute. Google’s survey discovered that 79% of organizations determine safety, governance, and AI operations as their main…
